Initial app

.gitignore

@@ -0,0 +1,3 @@
+.venv
+.env
+token.txt

app.py

@@ -0,0 +1,267 @@
+import cv2
+import gradio as gr
+from roboflow import Roboflow
+import numpy as np
+from PIL import Image
+import os
+from dotenv import load_dotenv
+import tempfile
+import time
+import random
+
+load_dotenv()
+
+# Initialize Roboflow
+api_key = os.getenv("ROBOFLOW_API_KEY")
+if not api_key:
+    raise ValueError("ROBOFLOW_API_KEY not found in environment variables.")
+roboflow = Roboflow(api_key=api_key)
+
+workspace = roboflow.workspace(
+    os.getenv("ROBOFLOW_WORKSPACE", "wayupuk-sommuang"))
+project = workspace.project(
+    os.getenv("ROBOFLOW_PROJECT", "face-detection-vswnd"))
+version = project.version(os.getenv("ROBOFLOW_VERSION", 11))
+
+model = version.model
+
+# Define emoji images with a fallback
+emoji_images = {}
+emoji_paths = {
+    "cat": 'images/cat_face.png',
+    "censored": 'images/censored.png',
+    "expressionless": 'images/expressionless.png',
+    "monkey": 'images/monky_see_no_evil.png'
+}
+for key, path in emoji_paths.items():
+    if os.path.exists(path):
+        emoji = cv2.imread(path, cv2.IMREAD_UNCHANGED)
+        if emoji is not None:
+            emoji_images[key] = emoji
+        else:
+            print(f"Failed to load emoji: {path}")
+    else:
+        print(f"Emoji file not found: {path}")
+
+# Fallback emoji (yellow square with transparency)
+if not emoji_images:
+    print("No emojis loaded. Using fallback.")
+    fallback_emoji = np.zeros((50, 50, 4), dtype=np.uint8)
+    fallback_emoji[..., :3] = (255, 255, 0)  # Yellow
+    fallback_emoji[..., 3] = 255  # Fully opaque
+    emoji_images["fallback"] = fallback_emoji
+
+# Dictionary to track faces and their assigned emojis
+face_to_emoji = {}
+emoji_keys = list(emoji_images.keys())
+
+
+def trim_transparency(image):
+    if image.shape[2] != 4:
+        return image
+
+    alpha = image[..., 3]
+    non_zero = cv2.findNonZero(alpha)
+    if non_zero is None:
+        return image
+
+    x, y, w, h = cv2.boundingRect(non_zero)
+    cropped = image[y:y+h, x:x+w]
+    return cropped
+
+
+def resize_emoji_to_fit(emoji, target_width, min_height):
+    emoji = trim_transparency(emoji)
+    height, width = emoji.shape[:2]
+    aspect_ratio = width / height
+
+    new_width = int(target_width)
+    new_height = int(new_width / aspect_ratio)
+
+    if new_height < min_height:
+        scale_factor = min_height / new_height
+        new_height = int(new_height * scale_factor)
+        new_width = int(new_width * scale_factor)
+
+    return cv2.resize(emoji, (new_width, new_height), interpolation=cv2.INTER_AREA)
+
+
+def match_faces(current_faces, threshold=50):
+    global face_to_emoji
+    new_face_to_emoji = {}
+    matched_faces = []
+
+    for face in current_faces:
+        center_x, center_y = face['x'], face['y']
+        face_id = None
+
+        for old_id, (old_x, old_y, emoji_key) in face_to_emoji.items():
+            distance = np.sqrt((center_x - old_x)**2 + (center_y - old_y)**2)
+            if distance < threshold:
+                face_id = old_id
+                break
+
+        if face_id is None:
+            face_id = f"face_{len(new_face_to_emoji)}"
+            emoji_key = random.choice(emoji_keys)
+            new_face_to_emoji[face_id] = (center_x, center_y, emoji_key)
+        else:
+            _, _, emoji_key = face_to_emoji[face_id]
+            new_face_to_emoji[face_id] = (center_x, center_y, emoji_key)
+
+        matched_faces.append((face, face_id, emoji_key))
+
+    face_to_emoji = new_face_to_emoji
+    return matched_faces
+
+
+def overlay_emoji_on_faces(frame, last_faces=None):
+    frame_height, frame_width = frame.shape[:2]
+    if last_faces is None:
+        return frame
+
+    if not emoji_keys:
+        return frame
+
+    matched_faces = match_faces(last_faces)
+
+    for face, face_id, emoji_key in matched_faces:
+        center_x, center_y = face['x'], face['y']
+        width, height = face['width'], face['height']
+
+        # Extend the bounding box height to include the mouth and chin
+        height_scale = 1.8
+        extended_height = height * height_scale
+        extended_center_y = center_y + (height * (height_scale - 1) / 2)
+
+        # Use the assigned emoji
+        emoji = emoji_images[emoji_key]
+
+        # Resize emoji to match the width and ensure minimum height
+        emoji_resized = resize_emoji_to_fit(emoji, width, extended_height)
+
+        # Get emoji dimensions
+        emoji_height, emoji_width = emoji_resized.shape[:2]
+
+        # Calculate top-left corner to center the emoji vertically on the extended bounding box
+        face_top = int(extended_center_y - extended_height / 2)
+        face_bottom = int(extended_center_y + extended_height / 2)
+        emoji_y_center = face_top + (face_bottom - face_top) // 2
+
+        # Adjust the y position with a reduced dynamic offset
+        offset_factor = max(0.1, 0.5 - height / 300)  # Reduced offset
+        vertical_offset = int(height * offset_factor)
+        x = int(center_x - emoji_width / 2)
+        y = int(emoji_y_center - emoji_height / 2 + vertical_offset)
+
+        # Ensure coordinates are within frame bounds
+        x = max(0, x)
+        y = max(0, y)
+        if x + emoji_width > frame_width:
+            emoji_width = frame_width - x
+        if y + emoji_height > frame_height:
+            emoji_height = frame_height - y
+
+        if emoji_width <= 0 or emoji_height <= 0:
+            continue
+
+        # Resize emoji again to fit the adjusted region
+        emoji_resized = cv2.resize(emoji_resized, (int(emoji_width), int(
+            emoji_height)), interpolation=cv2.INTER_AREA)
+
+        # Check if the emoji has an alpha channel
+        if emoji_resized.shape[2] == 4:
+            emoji_rgb = emoji_resized[..., :3]
+            alpha = emoji_resized[..., 3] / 255.0
+
+            frame_region = frame[y:y+int(emoji_height), x:x+int(emoji_width)]
+            if frame_region.shape[2] != 3:
+                continue
+
+            if frame_region.shape[:2] != emoji_rgb.shape[:2]:
+                emoji_rgb = cv2.resize(
+                    emoji_rgb, (frame_region.shape[1], frame_region.shape[0]), interpolation=cv2.INTER_AREA)
+                alpha = cv2.resize(
+                    alpha, (frame_region.shape[1], frame_region.shape[0]), interpolation=cv2.INTER_AREA)
+
+            alpha_3d = alpha[..., np.newaxis]
+            blended = (frame_region * (1 - alpha_3d) +
+                       emoji_rgb * alpha_3d).astype(np.uint8)
+            frame[y:y+int(emoji_height), x:x+int(emoji_width)] = blended
+        else:
+            frame[y:y+int(emoji_height), x:x+int(emoji_width)
+                  ] = emoji_resized[..., :3]
+
+        # Debug output
+        print(f"Face ID: {face_id}, Center: ({center_x}, {center_y}), Size: {width}x{height}, Extended height: {extended_height}, Emoji at ({x}, {y}), Size: {emoji_width}x{emoji_height}, Emoji used: {emoji_key}, Vertical offset: {vertical_offset}")
+
+    return frame
+
+
+def detect_faces(frame):
+    pil_image = Image.fromarray(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB))
+    with tempfile.NamedTemporaryFile(delete=False, suffix='.jpg') as temp_file:
+        temp_file_path = temp_file.name
+        pil_image.save(temp_file, format='JPEG')
+
+    try:
+        predictions = model.predict(
+            temp_file_path, confidence=40, overlap=30).json()
+        return predictions['predictions']
+    finally:
+        if os.path.exists(temp_file_path):
+            os.unlink(temp_file_path)
+
+
+def webcam_interface():
+    capture = cv2.VideoCapture(0)
+    if not capture.isOpened():
+        print("Error: Could not open webcam.")
+        return
+
+    capture.set(cv2.CAP_PROP_FRAME_WIDTH, 640)
+    capture.set(cv2.CAP_PROP_FRAME_HEIGHT, 480)
+
+    frame_count = 0
+    skip_frames = 10
+    last_faces = None
+    last_detection_time = 0
+    min_detection_interval = 0.2
+
+    try:
+        while True:
+            ret, frame = capture.read()
+            if not ret:
+                print("Error: Could not read frame.")
+                break
+
+            frame_count += 1
+            current_time = time.time()
+
+            if (frame_count % skip_frames == 0 or
+                    (current_time - last_detection_time) >= min_detection_interval):
+                last_faces = detect_faces(frame)
+                last_detection_time = current_time
+
+            frame = overlay_emoji_on_faces(frame, last_faces)
+
+            frame_rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
+            pil_img = Image.fromarray(frame_rgb)
+            yield pil_img
+
+    finally:
+        capture.release()
+
+
+# Gradio Interface
+iface = gr.Interface(
+    fn=webcam_interface,
+    inputs=[],
+    outputs=gr.Image(streaming=True),
+    live=True,
+    title="Face Detection with Emoji Overlay",
+    description="This app detects faces in the camera feed and overlays emojis."
+)
+
+if __name__ == "__main__":
+    iface.launch()

requirements.txt

@@ -0,0 +1,67 @@
+aiofiles==24.1.0
+annotated-types==0.7.0
+anyio==4.9.0
+certifi==2025.4.26
+charset-normalizer==3.4.1
+click==8.1.8
+contourpy==1.3.2
+cycler==0.12.1
+exceptiongroup==1.2.2
+fastapi==0.115.12
+ffmpy==0.5.0
+filelock==3.18.0
+filetype==1.2.0
+fonttools==4.57.0
+fsspec==2025.3.2
+gradio==5.27.0
+gradio_client==1.9.0
+groovy==0.1.2
+h11==0.16.0
+httpcore==1.0.9
+httpx==0.28.1
+huggingface-hub==0.30.2
+idna==3.7
+Jinja2==3.1.6
+kiwisolver==1.4.8
+markdown-it-py==3.0.0
+MarkupSafe==3.0.2
+matplotlib==3.10.1
+mdurl==0.1.2
+numpy==2.2.5
+opencv-python==4.11.0.86
+opencv-python-headless==4.10.0.84
+orjson==3.10.16
+packaging==25.0
+pandas==2.2.3
+pillow==11.2.1
+pillow_heif==0.22.0
+pydantic==2.11.3
+pydantic_core==2.33.1
+pydub==0.25.1
+Pygments==2.19.1
+pyparsing==3.2.3
+python-dateutil==2.9.0.post0
+python-dotenv==1.1.0
+python-multipart==0.0.20
+pytz==2025.2
+PyYAML==6.0.2
+requests==2.32.3
+requests-toolbelt==1.0.0
+rich==14.0.0
+roboflow==1.1.61
+ruff==0.11.7
+safehttpx==0.1.6
+semantic-version==2.10.0
+shellingham==1.5.4
+six==1.17.0
+sniffio==1.3.1
+starlette==0.46.2
+tomlkit==0.13.2
+tqdm==4.67.1
+typer==0.15.2
+typing-inspection==0.4.0
+typing_extensions==4.13.2
+tzdata==2025.2
+urllib3==2.4.0
+uvicorn==0.34.2
+websockets==15.0.1